Topic 4: 7 Layers of context for better AI prompts
Contextual prompting
A good prompt tells AI what you want. A stronger prompt also helps AI understand the situation around what you want.
That is contextual prompting.
For Product Managers and Business Analysts, context matters because our work rarely happens in isolation. A requirement belongs to a product. A user story belongs to a user journey. A recommendation needs to support a business goal. A roadmap needs to reflect priorities and constraints.
The more relevant context AI understands, the better it can support the work you are doing.
Better context, better output
Imagine asking AI:
“Create a roadmap.”
AI can certainly create something, but it has almost nothing to work with. The result is likely to be generic.
Now add some context:
“Create a 6-month roadmap for our B2B SaaS product focused on improving customer onboarding.”
Already, the AI understands much more.
Now imagine also providing your existing roadmap template, customer research, product goals, priorities and an example of a roadmap your organisation uses.
The AI is no longer guessing as much. It has information it can use to produce something that is more relevant to your actual situation.
This is the basic principle behind contextual prompting.
The 7 Layers of context
You do not need to put every possible piece of information into every prompt.
Instead, think about the different layers of context that could help AI understand your task.
1. User intent
What are you actually trying to achieve?
Be clear about the purpose behind the request.
For example, are you trying to understand a problem, prepare for a workshop, create a roadmap, improve a requirement or make a product decision?
2. Business context
Give AI the relevant background.
This might include:
- the product or service
- business goals
- current situation
- industry
- organisational environment
- relevant business information
This helps AI understand the bigger picture behind the task.
3. Audience
Who is the output for?
A recommendation for a senior leadership team will probably need to be different from something prepared for developers, customers or a product team.
Knowing the audience helps AI adjust the level of detail, language and focus.
4. Role / Persona
Sometimes it helps to tell AI what perspective you want it to adopt.
For example:
“Act as an experienced Business Analyst reviewing these requirements.”
Or:
“Review this from the perspective of a Product Manager responsible for customer onboarding.”
The role gives AI a perspective for approaching the task.
5. Constraints
Tell AI about important boundaries, assumptions, rules or limitations.
For example:
“Do not suggest changes that require a new technology platform.”
Or:
“Only use evidence from the research provided.”
Constraints help keep the response relevant to the real situation you are working within.
6. Format
Tell AI how you want the output structured.
You might ask for:
- a table
- a report
- user stories
- an OKR
- a roadmap
- a checklist
- recommendations
- a presentation outline
The right format makes the response easier to use in your actual work.
7. Examples
Examples can be one of the most useful forms of context.
You could give AI:
- a previous document
- an existing template
- a good user story
- an approved roadmap
- a past output
- a sample of the style you want
Instead of describing everything in words, you are showing AI what good looks like.
In-Context learning in action
This is where contextual prompting becomes particularly useful.
Imagine you want AI to create a product roadmap.
You could start with:
“Create a roadmap.”
That gives AI very little context.
A better approach would be to provide your roadmap template and say:
“Use the attached roadmap template and create a 6-month roadmap for our product.”
Now AI has a structure to follow.
You could go one step further by providing the template plus an example of a completed roadmap.
AI can then understand both the structure you want and how that structure has been used previously.
This is an example of in-context learning. You give AI relevant examples, templates or information within the conversation so it has something useful to learn from and follow.
Context can come from more than your prompt
Context does not always have to be typed manually into a prompt.
You can provide AI with useful context through:
- documents and reports
- templates and frameworks
- previous examples and outputs
- data and analytics
- roles and personas
The important question is:
What does AI need to know to do this job well?
Give it that information rather than expecting it to guess.
DSM Tip
Better context leads to better conversations.
You do not need all seven layers every time. Start with the context that matters for the task. For more important or complex work, give AI richer and more structured context so it can produce something genuinely useful.
